Single-shot 3D multi-person pose estimation in complex images - Sorbonne Université Access content directly
Journal Articles Pattern Recognition Year : 2021

Single-shot 3D multi-person pose estimation in complex images


In this paper, we propose a new single shot method for multi-person 3D human pose estimation in complex images. The model jointly learns to locate the human joints in the image, to estimate their 3D coordinates and to group these predictions into full human skeletons. The proposed method deals with a variable number of people and does not need bounding boxes to estimate the 3D poses. It leverages and extends the Stacked Hourglass Network and its multi-scale feature learning to manage multi-person situations. Thus, we exploit a robust 3D human pose formulation to fully describe several 3D human poses even in case of strong occlusions or crops. Then, joint grouping and human pose estimation for an arbitrary number of people are performed using the asso-ciative embedding method. Our approach significantly outperforms the state of the art on the challenging CMU Panoptic. Furthermore, it leads to good results on the complex and synthetic images from the newly proposed JTA Dataset.
Fichier principal
Vignette du fichier
Abd1.pdf (2.67 Mo) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-02926239 , version 1 (01-09-2020)



Abdallah Benzine, Bertrand Luvison, Quoc-Cuong Pham, Catherine Achard. Single-shot 3D multi-person pose estimation in complex images. Pattern Recognition, 2021, ⟨10.1016/j.patcog.2020.107534⟩. ⟨hal-02926239⟩
121 View
182 Download



Gmail Mastodon Facebook X LinkedIn More